[GRASS-dev] Handling of Python scripts on MS Windows

[Benjamin Ducke]

On 04/02/14 16:42, Vaclav Petras wrote:
> 
> 
> 
> On Tue, Feb 4, 2014 at 10:14 AM, Benjamin Ducke <benducke at fastmail.fm
> <mailto:benducke at fastmail.fm>> wrote:
> 
>     On 04/02/14 15:09, Vaclav Petras wrote:
>     >
>     > On Tue, Feb 4, 2014 at 8:25 AM, Benjamin Ducke
>     <benducke at fastmail.fm <mailto:benducke at fastmail.fm>
>     > <mailto:benducke at fastmail.fm <mailto:benducke at fastmail.fm>>> wrote:
>     >
>     >     But with the
>     >     latest Python developments, it will be a real challenge to
>     >     integrate GRASS 7 in the same way that we could integrate
>     >     GRASS 6 into a "host" application (i.e. without messing
>     >     around with any system settings).
>     >
>     >
>     > Can you be more specific? It sounds like something we should be
>     concern
>     > about. We should consider how GRASS 7 will be integrated into
>     other GIS
>     > packages when Python scripts are called this or that way.
> 
>     Sure. At this moment, we (i.e. the gvSIG CE dev team) distribute
>     GRASS 6 binaries along with gvSIG CE. On Windows, we also package
>     MSYS with "sh.exe" for the scripted modules.
> 
>     So we can deliever gvSIG CE + GRASS as a completely self-contained
>     (portable) package.
> 
>     Of course, we do not ship any GRASS GUI (or any of the d.* modules).
>     All GRASS modules are available seamlessly from the gvSIG CE GUI via a
>     Plug-in, similar to that in QGIS.
> 
>     Now, with the new Python-scripted modules in GRASS 7 this will
>     become more difficult, since Python adds a more heavy dependency.
>     In order to stay portable, we need to bundle a completely isolated
>     Python interpreter with gvSIG CE. And then we also need to make
>     sure that no GRASS Python module will call the system Python. The
>     latter is something that can only be taken care of on the GRASS
>     side.
> 
> 
> So, what would be the consequences for gvSIG, QGIS, OSGeo4W, ... if we
> decide to use system Python in (standalone) WinGRASS?

As far as gvSIG CE is concerned, there would be no consequences,
as long as it is possible to also make a GRASS distribution that
does not depend on system Python.

We ship our own, bundled GRASS binaries and users can just install
a WinGRASS (or any other GRASS version) in parallel. However,
in order to be able to bundle GRASS 7 in the future, none
of its modules should depend on system Python. If there were such
a dependency, then we would have to drop the Python-scripted
modules from the set of GRASS modules supported in gvSIG CE, as
we do not want to compromise our software's portability.

Best,

Ben


-- 
Dr. Benjamin Ducke, M.A.
{*} Geospatial Consultant
{*} GIS Developer

  benducke at fastmail.fm